NVIDIA B100 SXM Tensor Core GPU

Brand: NVIDIA | Category: GPUs

SKU: B100-SXM-192GB | Part #: B100-SXM-192GB | MPN: B100-SXM-192GB

Contact for Pricing — Request a Quote

The NVIDIA B100 SXM Tensor Core GPU delivers exceptional compute performance at FP32 with 5,456 TFLOPS, TF32 tensor operations reaching 10,912 TFLOPS, and FP64 calculations at 2,728 TFLOPS—three critical precision points that power modern AI inference and high-performance computing workloads. Built on NVIDIA's Blackwell architecture with CUDA Compute Capability 9.0, the B100-SXM-192GB combines 192 GB of HBM3e memory with industry-leading 960 GB/s memory bandwidth, enabling rapid data movement for dense matrix operations and transformer model deployments.

This server-class accelerator is engineered for AI infrastructure teams and enterprise data center operators who demand maximum throughput in space-constrained environments. The SXM form factor integrates seamlessly into standard GPU-compute servers via its 12V-2x6 / 16-pin connector, consuming 700W of power while maintaining stable operation across the 0–85°C temperature range. NVIDIA backs the B100 with a 3-year manufacturer warranty, with extended coverage available for mission-critical deployments. Operating under CE and FCC certification, this GPU is production-ready for regulated environments and large-scale GPU clusters.

Key Specifications

• GPU Model: B100
• Part Number: B100-SXM-192GB
• Memory: 192 GB HBM3e
• Peak FP32 Performance: 5,456 TFLOPS
• Tensor Performance (TF32): 10,912 TFLOPS
• Memory Bandwidth: 960 GB/s
• Power Consumption: 700W
• Form Factor: SXM (Server Module)
• Power Connector: 12V-2x6 / 16-pin
• Warranty: 3-year manufacturer warranty

For specifications, availability, and bulk licensing options, contact Omnixon Global to submit your RFQ.